mozilla :: #perf

20 Apr 2017
09:56jmaher|ptoigoldan: ping
10:37igoldanjmaher|pto: pong
10:41jmaherigoldan: hi, lots of bugs and backfilling today
10:41igoldanjmaher: hi
10:41igoldanjmaher: I'm now looking again over the duplicate one
10:41igoldanjmaher: I didn't see any previous merge
10:42igoldanjmaher: I'm rechecking the data points
11:02igoldanjmaher: could you explain more on this duplicate issue?
11:03igoldanjmaher: the 2 offending changesets are not related
11:03igoldanjmaher: autoland does a later merge with central
11:04igoldanjmaher: from that merge, a very small regression is visible, probably from the changeset existing in inbound
11:11igoldanjmaher|bbiab: does it have to do with the stability of the new bloom tests?
11:28jmaherit is confusing, it can happen that we get a regression on each branch before they merge, but then I would expect to see a larger overall regression
11:28jmaherin this case it isn't larger
11:29jmaheralso can you update the tabpaint bug? the retriggers are in and it looks like:
11:29jmaherigoldan: ^ :)
11:30igoldanjmaher: looking on it
11:31jmaherigoldan: let me look at the bloom data in more detail
11:32igoldanjmaher: ok
11:34jmaherigoldan: I could be wrong
11:34jmaherI am still confused as to why we don't see a cumulative regression
11:34jmahermaybe this test is just too sensitive
11:35igoldanjmaher: yes; it did confuse me
11:35igoldanjmaher: should have looked more obvious
11:35jmaherigoldan: I suspect both issues are just noise
11:36jmaherlooking at:,1d83527507421790c3e598b6fde76955bce2467d,1,1%5D&series=%5Bmozilla-inbound,1d83527507421790c3e598b6fde76955bce2467d,0,1%5D&zoom=1492582961804.7017,1492682179000,667.3357995469331,792.2428627439591
11:36jmaherI see the regression, then a dip down temporarily, then again
11:36jmaherit is really confusing
11:45igoldanjmaher: at the tabpaint filed bug, should I add the assignee also?
11:45igoldanjmaher: I added the blocking bug
12:06jmaherigoldan|bbiab: please do not assign bugs to people, needinfo is best
12:06igoldanjmaher: got it
12:14jmaherigoldan: good job on the tabpaint bug
12:15igoldanjmaher: thanks
12:16igoldanjmaher: I'm about to do the try retriggers
12:16jmaherigoldan: for the bloom changes, I am thinking we are seeing real changes, but possibly there are other things at play here (timing issues, build issues, sensitive test issues)
12:18igoldanjmaher: do we have any other docs/extra description about the bloom test?
12:18igoldanjmaher: currently, this is an event propagation test for me
12:22jmaherthat is it, we should work with bholley, etc to look at this
12:36igoldanjmaher: what's the name of the test suite for bloom?
12:36igoldanjmaher: perf-reftest-e10s?
12:37igoldanjmaher: it doesn't appear on TryChooser
12:37igoldanjmaher: or this is part of integrating a new test suite
12:38igoldanjmaher: where we know about it, and when sure about it, we make it public
12:38igoldanjmaher: ?
12:43jmaherigoldan: yes, that is the test
12:43jmaherigoldan: it is new on try- out of curiousity why would you want to push to try with that test
12:44jmaherigoldan: you should be able to add -t perf-reftest and it should work
12:48igoldanjmaher: was raised for e10s
12:52igoldanjmaher: also, I cannot add non-e10s test data on the charts
12:52igoldanjmaher: why is that?
12:54igoldanjmaher: does this foresee the drop of explicit e10s, when it will become the default?
12:58jmaherigoldan: you should be able to add non-e10s
13:10igoldanjmaher: when you push a backout on Try, do you use the qbackout extension for hg?
13:32chuttensmaug: Oh wow, you can refuse review requests with an HTTP response code? That's amazing.
13:33chutten&quot;Bugzilla error: Olli Pettay [:smaug] <> is not currently accepting &#39;review&#39; requests. (HTTP 500, API Error 225))&quot;
13:33chutten(this is from mozreview)
13:33smaugI always use bugzilla UI so didn&#39;t know that
13:33smaugI&#39;ll reopen the queue later today or tomorrow morning
13:34chuttenI&#39;m impressed. Also, I&#39;m curious who I might ask who isn&#39;t you for things within PresShell::processEvent
13:34smaug( just trying to reduce the review throughput a bit )
13:34smaughmm, ::ProcessEvent
13:34chutten(I was looking for other names, but it&#39;s pretty much just you through the bottom of that function, isn&#39;t it. Which said to me you&#39;re probably really really busy)
13:35chuttensmaug: Er, :handleEventInternal, rather
13:35smaugmasayuki is co-owner of event handling
13:36chuttenAlrighty, I&#39;ll r?masayuki and stop bothering you :)
13:56igoldanrwood|bbl hi! are you online?
15:15chuttenYou know your computer is overpowered when you forget and leave the Gecko Profiler running...
21 Apr 2017
No messages
Last message: 8 days and 45 minutes ago